Partition Coefficients for Metals in Surface Water, Soil and Waste

8/10/2019 Partition Coefficients for Metals in Surface Water, Soil and Waste

1/93

EPA/600/R-05/074

July2005

PARTITIONCOEFFICIENTSFOR

METALS

INSURFACEWATER,SOIL,ANDWASTE

by

JerryD.Allison,1,2

TerryL.Allison,.2

1HydroGeoLogic,Inc

1155HerndonParkway,Suite900

Herndon,VA20170

2AllisonGeoscienceConsultants,Inc.

3920PerryLane

FloweryBranch,GA30542

WorkAssignmentManager: RobertB.Ambrose,Jr.,P.E.

ContractNo.68-C6-0020 EcosystemsResearchDivision

NationalExposureResearchLaboratory

960CollegeStationRoad

Athens,GA30605

U.S.EnvironmentalProtectionAgency

OfficeofResearchandDevelopment

Washington,DC20460


2/93

NOTICE

TheinformationinthisdocumenthasbeenfundedwhollybytheUnitedStates

EnvironmentalProtectionAgencyundercontract68-C6-0020,WorkAssignment2-01to

HydroGeoLogic,Inc.IthasbeensubjecttotheAgency'speerandadministrativereview,andit

hasbeenapprovedforpublicationasanEPAdocument.Mentionoftradenamesofcommercialproductsdoesnotconstituteendorsementorrecommendationforuse.

ii


3/93

FOREWORD

TheNationalExposureResearchLaboratoryEcosystemsResearchDivision(ERD)in

Athens,Georgia,conductsprocess,modeling,andfieldresearchtoassesstheexposurerisksof

humansandecosystemstobothchemicalandnon-chemicalstressors.Thisresearchprovides

data,modeling,tools,andtechnicalsupporttoEPAProgramandRegionalOffices,stateandlocalgovernments,andothercustomers,enablingachievementofAgencyandORDstrategic

goalsfortheprotectionofhumanhealthandtheenvironment.

ERDresearchincludesstudiesofthebehaviorofcontaminants,nutrients,andbiotain

environmentalsystems,andthedevelopmentofmathematicalmodelstoassesstheresponseof

aquaticsystems,watersheds,andlandscapestostressesfromnaturalandanthropogenicsources.

ERDfieldandlaboratorystudiessupportprocessresearch,modeldevelopment,testingand

validation,andthecharacterizationofvariabilityandpredictionuncertainty.

Leading-edgecomputationaltechnologiesaredevelopedtointegratecorescience

researchresultsintomulti-media(air,surfacewater,groundwater,soil,sediment,biota),multi-stressor,andmulti-scale(organism,population,community,ecosystem;fieldsite,watershed,

regional,national,global)modelingsystemsthatprovidepredictivecapabilitiesforcomplex

environmentalexposurescenariosfacebytheAgency.

ExposuremodelsaredistributedandsupportedviatheEPACenterforExposure

AssessmentModeling(CEAM)(www.epa.gov/athens/ceampubl),theWatershedandWater

QualityModelTechnicalSupportCenter(www.epa.gov/athens/wwqtsc),andthroughaccessto

Internettools(www.epa.gov/athens/onsite).

ThisresearchprojectisacomponentoftheERDhazardouswasteresearchprogram,

whichseekstobetterunderstandtheenvironmentalcycling,exposure,andriskarisingfromthereleaseoforganicandinorganicpollutantsfromtreatmentfacilities.Inthisproject,metal

partitioncoefficientsweredevelopedforthewatershed,surfacewater,andsourcemodelsused

intheMultimedia,Multi-pathway,Multi-receptorExposureandRiskAssessment(3MRA)

technology.Knowledgeanddatagainedinthisevaluationwillbeusedtoimproveexposure

andriskanalysiscapabilitiesforheavymetalsevaluatedbythe3MRAandothermodelsused

byEPAinvariousregulatoryprograms.

EricJ.Weber,Ph.D.,ActingDirectorEcosystems

ResearchDivision

Athens,Georgia

iii


4/93

ABSTRACT

Thisreportpresentsmetalpartitioncoefficientsforthesurfacewaterpathwayandfor

thesourcemodelusedintheMultimedia,Multi-pathway,Multi-receptorExposureandRisk

Assessment(3MRA)technologyunderdevelopmentbytheU.S.EnvironmentalProtectionAgency.Partitioncoefficientsvaluesarepresentedforpartitioningbetweensoilandwater;

partitioningbetweenthesuspendedsedimentloadandthewaterinstreams,rivers,andlakes;

partitioningbetweenriverineorlacustrinesedimentanditsporewater;andpartitioningbetween

dissolvedorganiccarbon(DOC)andtheinorganicsolutionspeciesinthewaterofstreams,

rivers,andlakes. Somepartitioncoefficientsarealsopresentedtorepresentmetalpartitioning

betweenthesolidphaseofwasteanditsassociatedleachate.Partitioncoefficientsare

presentedforantimony(Sb),arsenic(As),barium(Ba),beryllium(Be),cadmium(Cd),

chromium(Cr),cobalt(Co),copper(Cu),lead(Pb),molybdenum(Mo),mercury(Hg),

methylatedmercury(CH3Hg),nickel(Ni),selenium(Se),silver(Ag),thallium(Tl),tin(Sn),

vanadium(V),andzinc(Zn).

Atwo-phaseapproachwasusedindevelopingtheneededpartitioncoefficients.Inthe

first-phase,aliteraturesurveywasperformedtodeterminetherangeandstatisticaldistribution

ofvaluesthathavebeenobservedinfieldstudies. Thisincludedthecollectionofpublished

partitioncoefficientsforanyofthemetalsinanyoftheenvironmentalmediaofinterest,orour

estimationofpartitioncoefficientsfromreportedmetalconcentrationdatawhenfeasible.In

thesecond-phaseeffort,statisticalmethods,geochemicalspeciationmodeling,andexpert

judgementwereusedtoprovidereasonableestimatesofthosepartitioncoefficientsnot

obtainedfromourliteraturesearchanddataprocessing.Thereportconcludeswithadiscussion

ofthemanysourcesofuncertaintyinthereportedmetalpartitioncoefficients.

iv


5/93

TABLEOFCONTENTS

page

1.0 INTRODUCTIONANDBACKGROUND. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-1

2.0 LITERATURESURVEYFORMETALPARTITIONCOEFFICIENTS . . . . . . . . 2-12.1 SELECTIONCRITERIAFORPARTITIONCOEFFICIENTS . . . . . . . . . . 2-3

2.2 RESULTSOFTHELITERATURESURVEY. . . . . . . . . . . . . . . . . . . . . . . 2-3

3.0 ANALYSISOFRETRIEVEDDATAANDDEVELOPMENTOF

PARTITIONCOEFFICIENTVALUES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-1

3.1 DEVELOPMENTOFPARTITIONCOEFFICIENTSFOR

NATURALMEDIA. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-1

3.1.1 EstimationfromRegressionEquationsBasedonLiteratureData. . . 3-2

3.1.2 EstimationFromGeochemicalSpeciationModeling . . . . . . . . . . . . 3-3

3.1.3 EstimationfromExpertJudgement . . . . . . . . . . . . . . . . . . . . . . . . . . 3-6

3.2 DEVELOPMENTOFPARTITIONCOEFFICIENTSFORWASTESYSTEMS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-16

3.2.1 EstimationfromAnalysisofDataPresentedintheLiterature. . . . . 3-17

3.2.2 EstimationfromGeochemicalSpeciationModeling. . . . . . . . . . . . 3-19

4.0 DISCUSSIONOFRESULTSANDSOURCESOFUNCERTAINTY . . . . . . . . . . 4-1

5.0 REFERENCES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-1

APPENDICES

APPENDIXA

APPENDIXB

APPENDIXC

APPENDIXD

APPENDIXE

METALPARTITIONCOEFFICIENTSUSEDINSOMERECENTU.S.EPARISKASSESSMENTS

SCATTERPLOTSFORLINEARREGRESSIONSUSEDTO

ESTIMATEMEANLOGKINNATURALMEDIA

EXAMPLEINPUTFILEFORTHEMINTEQA2MODEL

USEDTOESTIMATEMETALPARTITIONINGIN

SOIL/SOILWATERSYSTEMS


USEDTOESTIMATEMETALPARTITIONINGTODOC


USEDTOESTIMATEMETALPARTITIONINGIN

WASTEMANAGEMENTSYSTEMS

v


6/93

LISTOFTABLES

page

Table1 Partitioncoefficients(logKdinL/kg)fromtheliteraturesearch.. . . . . . . . . 2-5

Table2 LinearregressionequationsusedtoestimatemeanlogKdvalues(L/kg)innaturalmedia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-3

Table3 Metalpartitioncoefficients(logKd)inkg/Lforsoil/soilwater . . . . . . . . . . 3-8

Table4 Metalpartitioncoefficients(logKd)inkg/Lforsediment/porewater . . . . . 3-10

Table5 Metalpartitioncoefficients(logKd)inkg/Lforsuspendedmatter/water . . 3-12

Table6 Metalpartitioncoefficients(logKd)inkg/Lforpartitioningbetween

DOCandinorganicsolutionspecies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-14

Table7 Effectivemetalpartitioncoefficientsbasedonreportedsolidphaseandsolution

phasemetalsconcentrationsfromleachtestsreportedintheliterature. . . . 3-18

Table8 ImportantparametersandconstituentconcentrationsusedinMINTEQA2

modelingoflandfillsintheacetogenicandmethanogenicstagesand

MSWIandCKDmonofills . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-20Table9 Estimatedrangeinlogpartitioncoefficients(L/kg)inwasteforselected

metalsdeterminedfromMINTEQA2modeling . . . . . . . . . . . . . . . . . . . . . 3-22

vi


7/93

LISTOFFIGURES

page

FigureB-1 DatausedtodevelopregressionequationtopredictsedimentKd

fromsoilKd . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-1FigureB-2 DatausedtodevelopregressionequationtopredictsedimentKd

fromsuspendedmatterKd . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-1

FigureB-3 DatausedtodevelopregressionequationtopredictsoilKdfromsuspendedmatterKd(andviceversa) . . . . . . . . . . . . . . . . . . . . . . . . . . B-2

FigureB-4 DatausedtodevelopregressionequationtopredictwasteKdfromsoilKd . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-2

vii


8/93

1.0 INTRODUCTIONANDBACKGROUND

Thepurposeofthisstudywastodevelopmetalpartitioncoefficientsforthesurfacewater

pathwayandforthesourcemodelusedintheMultimedia,Multi-pathway,Multi-receptor

ExposureandRiskAssessment(3MRA)technologyunderdevelopmentbytheU.S.

EnvironmentalProtectionAgency.The3MRAtechnologyprovidesforscreening-levelhumanandecologicalriskassessmentsforchronicexposuretochemicalsreleasedfromland-based

wastemanagementunits(WMUs)managedundertheHazardousWasteIdentificationRule

(HWIR). Themultimedia3MRAmodelincludesasurfacewaterpathwaymodelthatrequires

thepartitioncoefficientforeachmetaltobemodeled.

Innaturalmedia,metalcontaminantsundergoreactionswithligandsinwaterandwithsurface

sitesonthesolidmaterialswithwhichthewaterisincontact.Reactionsinwhichthemetalis

boundtothesolidmatrixarereferredtoassorptionreactionsandmetalthatisboundtothe

solidissaidtobesorbed.Themetalpartitioncoefficient(Kd;alsoknownasthesorption

distributioncoefficient)istheratioofsorbedmetalconcentration(expressedinmgmetalper

kgsorbingmaterial)tothedissolvedmetalconcentration(expressedinmgmetalperLofsolution)atequilibrium.

(1)

Duringtransportofmetalsinsoilsandsurfacewatersystems,metalsorptiontothesolidmatrix

resultsinareductioninthedissolvedconcentrationofmetalandthisaffectstheoverallrateof

metaltransport. Thus,transportmodelssuchasthoseusedinvariouspathwaysinthe3MRA

incorporatethemetalKdintotheoverallretardationfactor(theratiooftheaveragelinear

particlevelocitytothevelocityofthatportionoftheplumewherethecontaminantisat50percentdilution). TheuseofKdin3MRAtransportmodelingimpliestheassumptionthatlocal

equilibriumbetweenthemetalsolutesandthesorbentsisattained. Thisimpliesthattherateof

sorptionreactionsisfastrelativetoadvective-dispersivetransportofthemetal.

Foraparticularmetal,Kdvaluesinsoilaredependentuponvariousgeochemicalcharacteristics

ofthesoilanditsporewater. LikewiseforsurfacewatersystemstheKdforaparticularmetal

dependsonthenatureofsuspendedsolidsorsedimentandkeygeochemicalparametersofthe

water. GeochemicalparametersthathavethegreatestinfluenceonthemagnitudeofKdinclude

thepHofthesystemandthenatureandconcentrationofsorbentsassociatedwiththesoilor

surfacewater. Inthesubsurfacebeneathawastemanagementfacility,theconcentrationof

leachateconstituentsmayalsoinfluencethemetalKdthroughcompetitionforsorptionsites.

ThemetalsofinterestinHWIRmodelingareantimony(Sb),arsenic(As),barium(Ba),

beryllium(Be),cadmium(Cd),chromium(Cr),cobalt(Co),copper(Cu),lead(Pb),

molybdenum(Mo),mercury(Hg),nickel(Ni),selenium(Se),silver(Ag),thallium(Tl),tin

1-1


9/93

(Sn),vanadium(V),andzinc(Zn).Methylatedmercury(CH3Hg)andcyanide(CN)arealsoof

interest. Inthesurfacewaterpathway,the3MRAincludesseveraltransportprocessesthat

requiremetalpartitioncoefficients:(1)Theoverlandtransportofmetalcontaminantsinrunoff

waterinthewatershedandtheconsequentpartitioningbetweensoilandwater;(2)partitioning

betweenthesuspendedsedimentloadandthewaterinstreams,rivers,andlakes;(3)

partitioningbetweenriverineorlacustrinesedimentanditsporewater;and(4)partitioningbetweendissolvedorganiccarbon(DOC)andtheinorganicsolutionspeciesinthewaterof

streams,rivers,andlakes.

The3MRAmodelingscenarioalsoincludesasourcemodelforvarioustypesofwaste

managementunitsthatalsorequirespartitioncoefficients.Forthesourcemodel,thepartition

coefficientsareusedtorepresenttheratioofcontaminantmassinthesolidphasetothatinthe

leachate(water)phase. Therearefivetypesofwastemanagementunitsforwhichthesource

modelrequirespartitioncoefficients:landapplicationunits,wastepiles,landfills,treatment

lagoons(surfaceimpoundments),andaeratedtanks.

Thisreportdescribesthetwo-phaseapproachusedindevelopingtheneededpartitioncoefficients. Inthepreferred(first-phase)methodofobtainingthecoefficients,aliterature

surveywasperformedtodeterminetherangeandstatisticaldistributionofvaluesthathave

beenobservedinfieldstudies.Thisincludesthecollectionofpublishedpartitioncoefficients

foranyofthemetalsinanyoftheenvironmentalmediaofinterest,orourestimationof

partitioncoefficientsfromreportedmetalconcentrationdatawhenfeasible.Thedataretrieved

intheliteraturesearchwererecordedinaspreadsheetalongwithassociatedgeochemical

parameters(suchaspH,sorbentconcentration,etc.)whenthesewerereported.Weanticipated

thattheliteraturesearchwouldnotsupplyneededpartitioncoefficientsforallofthemetalsin

alloftheenvironmentalmediaofinterest. Therefore,inthesecond-phaseeffort,statistical

methods,geochemicalspeciationmodeling,andexpertjudgementwereusedtoprovide

reasonableestimatesofthosepartitioncoefficientsnotobtainedfromourliteraturesearchanddataprocessing.

2.0 LITERATURESURVEYFORMETALPARTITIONCOEFFICIENTS

Aliteraturesurveywasconductedtoobtainpartitioncoefficientstodescribethepartitioningof

metalsbetweensoilandsoil-water,betweensuspendedparticulatematter(SPM)andsurface

water,betweensedimentandsediment-porewater,andbetweenDOCandthedissolved

inorganicphaseinnaturalwaters. Inaddition,partitioncoefficientsweresoughtfor

equilibriumpartitioningofmetalsbetweenwastematrixmaterialandtheassociatedaqueous

phaseinlandapplicationunits,wastepiles,landfills,treatmentlagoons,andaeratedtanks.The

literaturesurveyencompassedperiodicalscientificandengineeringmaterialsaswellassome

non-periodicals,includingbooksandtechnicalreportspublishedbytheU.S.EPAandother

2-1


10/93

governmentagencies. Electronicsearchesofthefollowingdatabaseswereincludedaspartof

theliteraturesurvey:

AcademicPressJournals(1995-present)

AGRICOLA(1970-present)

AnalyticalAbstracts(1980-present) AppliedScienceandTechnologyAbstracts

AquaticSciencesandFisheriesAbstractSet(1981-present)

CABAbstracts(1987-present)

CurrentContents(1992-present)

DissertationAbstracts(1981-present)

EcologyAbstracts(1982-present)

EISDigestofEnvironmentalImpactStatements(1985-present)

EITechIndex(1987-present)

EnvironmentalEngineeringAbstracts(1990-present)

GeneralScienceAbstracts(1984-present)

GEOBASE(1980-present) GEOREF(1785-present)

NationalTechnicalInformationService

PapersFirst(1993-present)

PeriodicalAbstracts(1986-present)

ToxicologyAbstracts(1982-present)

WaterResourcesAbstracts(1987-present)

Twosearchstringswereusedintheelectronicsearches:distributioncoefficientand

partitioncoefficient. Useofsuchgeneralstringshastheadvantageofgeneratingmany

citations,decreasingtheprobabilitythatrelevantarticleswillbemissed,butalsocarryinga

highlaborburdenbecauseeachcitationreturnedmustbeexaminedforusefuldata.Formetalsthatarenotaswellrepresentedinthepublishedliterature,evenmoregeneralsearchstrings

wereused,sometimeswithbooleanoperators(e.g.,bariumandsoil,seleniumand

partitioning).Theworkofidentifyingarticlescontainingusefuldatafromamongallthose

retrievedwasmadeeasierbyfirstreviewingthetitlestoeliminatethoseofobviousirrelevance,

thenreviewingtheabstracts,thatwereusuallyavailableon-line.Abstractsofcitationsthat

showedpromiseforprovidingpartitioncoefficientswereprintedandgivenacodeconsistingof

thefirsttwolettersoftheleadauthorslastnameandthelasttwodigitsoftheyearof

publication. Thecode,alongwiththefirstfewwordsofthearticletitle,wasenteredinalog

bookfortracking.Loggedarticleswerequicklyreviewedatlocaluniversityresearchlibraries,

andthosecontainingrelevantdatawerecopiedforamorethoroughreviewatouroffice.Most

ofthearticleswereobtainedfromtheUniversityofGeorgiaScienceLibraryortheGeorgia

InstituteofTechnologyLibrary.Aseachcopiedarticleorreportwasreviewed,asummary

pagecontainingtheassignedcodewasstapledtothefrontwithnotesindicatingthetypeofdata

foundinthepaperandthelocation(pagenumber,tablenumber,etc.)ofusefuldata.Partition

2-2


11/93

coefficientsandotherdatafromthearticleswerethenenteredintoanEXCEL97spreadsheet

forcompilationandanalysis.

2.1 SELECTIONCRITERIAFORPARTITIONCOEFFICIENTS

Thefollowingcriteriaandguidelineswerefollowedintheselectionofpartitioncoefficient

valuesfromjournalarticles.Valueswereacceptedfromstudiescharacterizedby:

C Useofwholenaturalmediafordeterminationofpartitioncoefficientsinnatural

mediasystems(e.g.,rejectvaluesfromstudiesusingpuremineralphasesortreated

soils)

C Insoilsystems,useofanextractanthavinglowionicstrength(#0.1M);insurface

watersystems,lowsalinity(freshwaterpreferred,salinityupto10partsperthousand

acceptable)

C Useoflowtotalmetalconcentrations(i.e.,ifcoefficientsweredeterminedatmultiple

totalmetalconcentrations,choosethecoefficientcorrespondingtothelowestconcentrationwhereKdislesslikelytodependonmetalconcentration)

C pHvaluesinthenaturalrange(4to10)

C Noorganicchelatesintheextractant(e.g.,EDTA)

C Onepartitioncoefficientpersystemstudied

C Wheremultiplepartitioncoefficientsarepresentedforasystemduetoexperimental

variationofpHorotherparameters,selectthepartitioncoefficientcorrespondingtothe

conditionsmostcloselyapproximatingnaturalconditions.

C Batchleachingtests(preferredovercolumntestsifbothareavailableforthesamestudy

andsoil,butcolumntestsacceptable).

Thegeochemicalparametersmostlikelytoinfluencethepartitioncoefficientwereenteredinthespreadsheetalongwithreportedorcalculatedcoefficientsifsuchwerespecifiedinthe

sourcearticleorreport.ExamplesoftheseparametersarepH,totalconcentrationsofmetalin

solutionandsorbed,andconcentrationsofimportantmetalcomplexingagents(including

DOC),andweightfractionofparticulateorganicmatterandothersorbingmaterials.Physical

parametersnecessarytoconvertsorbedconcentration(mg/kg)overdissolvedconcentration

(mg/L)topartitioncoefficientsinlitersperkilogram(L/kg),i.e.,porosity,watercontent,and

bulkdensity,werealsorecordedwhenreportedinthearticles.Equationsandrelationships

presentedinjournalarticlesthatpresentKdasafunctionofpHorotherparameterswere

recordedinaremarkfieldinthespreadsheet.

2.2 RESULTSOFTHELITERATURESURVEY

Approximately245articlesandreportswerecopiedandreviewed.Atotalof1170individual

Kdvalueswereobtainedfromthesesources,eitherdirectlyorcalculatedfromreportedmedia

concentrations. ThistotaldoesnotincludemeanestimatedKdvaluesreportedinpreviously

publishedcompilationsofKdvalues(BaesandSharp,1983;Baesetal.,1984;Coughtreyetal.,

2-3


12/93

1985;Thibaultetal.,1990). (Thedatafromthesepreviouscompilationswererecordedinthe

spreadsheetandusedinguidingthefinalestimatesofappropriatecentraltendencyvaluesas

describedinSection3.1.3.) Approximately80%ofthe1170valuesweobtainedfromthe

literaturepertainedtothemetalsCd,Co,Cr,Cu,Hg,Ni,Pb,andZn. MoreKdswere

recoveredforCdthananyothermetal,followedcloselybyZn,Pb,andCu.Themost

frequentlyreportedtypeofKdwasthatforsuspendedmatterinstreams,riversandlakes.(Datapertainingtomarineenvironmentsweregenerallyrejected,butsomedatafromestuarieswere

includedifreportedascorrespondingtolowsalinity.)ThesecondmostfrequentlyreportedKdvaluespertainedtopartitioninginsoil.SuspendedmatterandsoilKdstogethertotaled68%of

thereporteddata. Table1showsthemedianandrangeofKdvaluesretrievedinourliterature

searchfornaturalmedia.(ValuesshownarelogKdvalues). Forsomecombinationsofmetal

andmediatype,toofewpartitioncoefficientswerefoundintheliteraturetostateamedianor

evenareasonablerange.Insomeofthesecases,meanormedianvalueswereavailablefrom

previouscompilationsofpartitioncoefficients. InTable1,blankspacesinthetablecorrespond

tonodatafound. Valuesinboldarefrompreviouscompilations.

Nodirectlyreportedpartitioncoefficientsforthewastesystemsofinterestwerediscoveredintheliteraturesurvey,andnoneareincludedinTable1.Therearemanyreasonsforwishingto

understandthebehaviorofmetalsinnaturalsystems.Therichliteratureofsoilscience,plant

nutrition,aquaticchemistry,geology,andtoxicologyareallexamplesofinvestigativeareasof

longstandingwheremetalpartitioncoefficientsarefrequentlyencountered.Theimpetusfor

researchwithregardtowastesystemsissignificantlydifferentfromthatofnaturalsystems.

Moreover,thebehaviorofmetalsinwastematerialsaretypicallystudiedandreportedpriorto

theirdisposalandconsequentmixingwithahostofothersubstancesfewstudieshave

focusedonthebehaviorofmetalswithinactualdisposalunitscontaininga(usuallyunknown)

mixtureofmaterials.Moststudiesinvolvingmetalconcentrationsinwasteareconcernedwith

predictingthemetalconcentrationinleachatebymeansofaphysicaltest(i.e.,aleachate

extractiontest).Section3.2presentsfurtherfindingswithregardtoleachtestsandappropriatemetalpartitioncoefficientsforwastesystems.

2-4


13/93

Table1

Partitioncoefficients(log KdinL/kg)fromtheliteraturesearch.

Medianvalueslistedinboldfacearefromapreviouscompilation.

Blankspacesrepresentinstancesforwhichnodatawasfoundortoo

fewvalueswerefoundtoprovidemeaningfulstatistics.

Metal Soil/Water

Suspended

Matter

/Water

Sediment/

Water DOC/Water

Ag

median

2.6 4.9 3.6

range 1.0-4.5 4.4-6.3 2.1-5.8

N 21 15

As

median

3.4 4.0 2.5

range 0.3-4.3 2.0-6.0 1.6-4.3

N 22 25 18

Ba

median

4.0

range 0.7-3.4 2.9-4.5

N 14

Be

median3.1 4.1

range 1.7-4.1 2.8-6.8

N 2 17

Cd

median

2.9 4.7 3.6 5.2

range 0.1-5.0 2.8-6.3 0.5-7.3 3.4-5.5

N 41 67 21 4

Co

median

2.1 4.7 3.3 4.5

range (-1.2)-4.1 3.2-6.3 2.9-3.6 2.9-4.8

N 11 29 3 2

2-5


14/93

Table1





Metal Soil/Water

Suspended

Matter

/Water

Sediment/

Water DOC/Water

Cr(III)

median

3.9 5.1 4.5

range 1.0-4.7 3.9-6.0

N 43 25

Cr(VI)

median

1.1

range (-0.7)-3.3

N 24

Cu

median

2.7 4.7 4.2 5.5

range 0.1-3.6 3.1-6.1 0.7-6.2 2.5-7.0

N 20 70 12 17

Hg

median3.8 5.3 4.9 5.3

range 2.2-5.8 4.2-6.9 3.8-6.0 5.3-5.6

N 17 35 2 3

CH3Hg

median

2.8 5.4 3.6

range 1.3-4.8 4.2-6.2 2.8-5.0

N 11 2 4

Mo

median

1.1 2.5

range (-0.2)-2.7

N 8

2-6


15/93

Table1





Metal Soil/Water

Suspended

Matter

/Water

Sediment/

Water DOC/Water

Ni

median

3.1 4.6 4.0 5.1

range 1.0-3.8 3.5-5.7 4.7-5.4

N 18 30 4

Pb

median

4.2 5.6 5.1 5.0

range 0.7-5.0 3.4-6.5 2.0-7.0 3.8-5.6

N 33 48 24 9

Sb

median

2.4 4.0

range 0.1-2.7 2.5-4.8 2.7-4.3

N 3

Se

median 1.0 3.6

range -0.3-2.4 3.1-4.7

N 23

Sn

median

2.9 5.6 4.7

range 2.1-4.0 4.9-6.3

N 3

Tl

median3.2

range 3.0-3.5

N 6

2-7


16/93

Table1





Metal Soil/Water

Suspended

Matter

/Water

Sediment/

Water DOC/Water

V

median

range 1.1-2.7

N

Zn

median

3.1 5.1 3.7 4.9

range (-1.0)-5.0 3.5-6.9 1.5-6.2 4.6-6.4

N 21 75 18 9

CN

median

3.0

range 0.7-3.6

N 3

PartitioncoefficientsusedinseveralrecentU.S.EPAriskassessmentsarepresentedinAppendixA.Becausetheoriginofthesedataisgenerallyunknown,theywerenotincludedin

thecollectionofKdvaluesappearingelsewhereinourspreadsheet,norweretheyincludedin

thestatistical summaryofKdvaluesobtainedfromtheliteratureasreportedherein.

2-8


17/93

3.0 ANALYSISOFRETRIEVEDDATAANDDEVELOPMENTOFPARTITION

COEFFICIENTVALUES

Thedatagatheredfrompublishedsourceswereinsufficienttoestablishareasonablerange

and/ormedianvalueforthepartitioncoefficientforallmetalsinallmedia-types.Therefore,

thesecondpartoftheeffortwasdirectedataugmentingthevaluesobtainedfromtheliteraturesoastoprovideareasonablerangeandcentraltendencyofKdforeachmetalineachmedia-

type. Statisticalanalysisofretrieveddata,geochemicalmodeling,andexpertjudgementwere

allusedtodevelopthesepartitioncoefficientvalues. Thenatureoftheavailabledatafor

naturalmediaandwastesystemswasdifferenttotheextentthatitseemedbesttoconsider

thesetwocategoriesseparately.

3.1 DEVELOPMENTOFPARTITIONCOEFFICIENTSINNATURALMEDIA

Inanalyzingthepartitioningdatacollectedfromtheliteratureforsoilandsurfacewater

systems,weattemptedtoidentifytheshapeoftheprobabilitydistributionforeachmetalin

eachmedium. Foraparticularmetalinaparticularmedium,thedegreetowhichtheliteraturesampleistrulyrepresentativeofthepopulationofmetalpartitioncoefficientsisdependenton

thenumberofsamplepoints,theactualvariabilityofimportantmediumpropertiesthat

influencepartitioning(pH,concentrationofsorbingphases,etc.),andhowwellthisvariability

isrepresentedinthesample. Insomecases,itwasnecessarytoeliminatedatapointsfromthe

literaturesampletoavoidobviousbias. Forexample,thesampleofliteratureKdvaluesfor

Cr(III)insoilincludedvaluesobtainedinapHtitrationofthreesoilssuchthateachofthethree

wasrepresentedbyeightdifferentKdvalues. Althoughtheyprovideinterestingdataonthe

dependenceofKdonpHinthesesoils,multiplemeasurementsfromthesamesoilandvalues

determinedatotherthantheambientsoilpHintroducebiasinthenaturalprobability

distributionofKd. Therefore,incaseswhereKdassociatedwithmultiplepHvalueswere

presented,theKdassociatedwiththepHvalueclosesttotheambientsoilpHwaschosen.IftheambientsoilpHwasnotspecified,thenasingleKdvaluewaspickedrandomlyfromamong

thosepresentedandtheotherKdvaluesforthesoilwerediscarded.Inthisfashion,thesample

ofliteraturedataforeachmetalandmedia-typewaseditedbeforeattemptingtoidentifythe

underlyingprobabilitydistributionforKd.

StatisticaltestswereperformedtodeterminetheshapeofthefrequencydistributionofKdfor

eachmetalandmedia-type.Thesetestsemployedwidelyrecognizedtechniquesavailablein

thestatisticalpackageAnalyze-It(version1.32),anadd-onmoduleforMicrosoftEXCEL97.

TheShapiro-WilktestandtheKolmogorov-Smirnovtestwereusedtotestthesamplesfor

normality.ApositivetestinShapiro-Wilkdoesnotensureanormaldistribution.Rather,it

providesameasureofconfidencethatthesampledataarenotinconsistentwithanormal

distribution. TheShapiro-Wilktestisageneraltestfornormality;itisnotnecessarytoknow

thepopulationmeanorstandarddeviation. TheKolmogorov-Smirnovtestwasusedwhen

resultsfromtheShapiro-Wilktestwerenegative. Inonlyafewcaseswerethedatasufficient

toidentifytheunderlyingdistributionwithanydegreeofcertainty.Manyofthesamplesets

(includingthemostcomplete(largest)samplesets),gaveapositivetestfornormalityafter

3-1


18/93

transformingtheavailabledatatologspace,suggestingthatthefrequencydistributionofthe

underlyingpopulationofKdvaluesforaparticularmetalinaparticularmediumismostlikely

log-normal.

Insomecases,thereweretoofewrepresentativedatapointsinthesampletohaveconfidencein

thedescriptivestatisticsofthedata. Inthesecases,threemethodswereusedtoaugmenttheavailabledatainestimatingthemean,standarddeviation,andminimumandmaximumKdvalues. Thethreemethodswere:estimationfromlinearregressionequationsdevelopedfrom

theliteraturesamples,estimationfromtheresultsofgeochemicalspeciationmodeling,and

estimationbyexpertjudgement. Eachmethodisdiscussedbelow.

3.1.1 EstimationfromRegressionEquationsBasedonLiteratureData

Ofthe13metalsforwhichliteraturedatawereretrievedcharacterizingKdinsoil,sediment,

andsuspendedmatter,12ofthemexhibitedaprogressionofdecreasingaffinityforsorption

materialintheordersuspendedmatter>sediment>soil. Inotherwords,comparisonofmean

KdvaluesforparticularmetalsshowedtheresultthatKd,SPM>Kd,Sediment>Kd,Soil. IntwoothercaseswhereatleasttwooftheKdtypescouldbecharacterizedfromtheliteraturedata,both

conformedtothissamepattern. Inaddition,asomewhatconsistentprogressioninKdmagnitudeformetalswithinthethreenaturalmediawasnoted.Forthebestrepresented

metals,thefollowingpatternsofdecreasingKdwereobserved(basedonorderingthemeanKdvaluesfromhighesttolowestmagnitudeforeachmedium):

Soils: Pb>Cr III>Hg>As>Zn=Ni>Cd>Cu>Ag>Co

Sediment: Pb>Hg>CrIII>Cu>Ni>Zn>Cd>Ag>Co>As

SPM: Pb>Hg>Cr III=Zn>Ag>Cu=Cd=Co>Ni>As

TherewassomeshufflingaboutoftheKdmagnitudeorderingamongthesemedia-types,asmightbeexpectedforadatasetthatisundoubtedlyincomplete. Themostobvious

inconsistencyintheprogressionofKdmagnitudeisforAs. Nevertheless,thesimilaritiesare

worthyofnote.Someaspectsoftheoveralltrendareinagreementwiththehard-softacid-base

(HSAB)conceptsofPearson(1963),however,PbandHghavegreateraffinitiesthanHSAB

predicts.Certainly,therearemultipleadsorptionsurfacespresentinallofthesematerials.The

consistencyofaffinityrelationshipsamongthesemetalssuggeststhatthedistributionofKdis

partlyduetocharacteristicsuniquetothemetalsthemselvesandpartlyduetocharacteristics

associatedwiththesorbingsurfaces.Regardlessofthereason,itappearsfeasibletoexploit

thesetrendstoprovideanestimateofKdforagivenmetalinonemediumifitsvalueinanother

mediumisavailable. Forexample,theliteraturedataprovidedareasonablenumberofKdvaluesinsoilsandsuspendedmatterfortheninemetalsAg,Cd,Co,Cr(III),Cu,Hg,Ni,Pb,

andZn. Foreachofthesemetals,themeanvalueofKdinsoilwasintheneighborhoodoftwo

ordersofmagnitudelessthanthemeanvalueinsuspendedmatter.Thistrendwas

characterizedmoreexactlybydevelopingalinearregressionequation.Theregressionequation

wasthenusedtoestimatemeanKdvaluesformetalsforwhichtheliteratureprovidedan

estimateofmeanKdinsoil,butnotinsuspendedmatter. Inasimilarmanner,linearregression

3-2


19/93

equationsweredevelopedtoestimatethemeanKdinsedimentfromtheliteratureestimateof

meanKdinsoilorsuspendedmatter,orthemeansoilKdfromthatinsedimentorsuspended

matter. Theregressionequationsweredevelopedfromcaseswheretheliteraturesurveydata

providedreasonableestimatesofthemeanKdforatleasttwoofthethreemedia.Themetals

usedindevelopingtheregressionequationsincludedcadmium,copper,zinc,andothermetals

thatwerebetterrepresentedintheliterature.ThedistributionofKdvaluesforaparticularmetalwasassumedtobelog-normalsothattheregressionequationswereactuallybasedon

meanlogKdandwereusedtopredictmeanlogKd. Thestandarddeviationwasestimatedfrom

themeanandminimumvaluesassumingtheminimumvaluerepresentstwostandard

deviationsfromthemean. Thestandarddeviationwasalsoestimatedusingthemeanand

maximumvaluesratherthanmeanandminimum. Thelargerofthetwoestimatesofstandard

deviationwasretainedasthefinalestimate.TheregressionequationsusedareshowninTable

2alongwiththenumberofobservationsuponwhicheachequationisbased,thecorrelation

coefficient(r2),andthe95%confidenceintervalfortheslopeandintercept.Scatterplots

showingtheregresseddatapointsandstraightlineregressionsareshowninAppendixB.

Table2LinearregressionequationsusedtoestimatemeanlogKdvalues(L/kg)innaturalmedia.

Usedto Independent slope intercept

Estimate Variable (+/-95%CI) (+/-95%CI) r2 N

meanlogKd meanlogKd 1.080 0.796 0.7 5

sediment soil (1.035) (3.190) 9

meanlogKd meanlogKd 1.418 -3.179 0.6 5

sediment suspended (1.923) (9.868) 5

matter

meanlogKd meanlogKd 0.380 3.889 0.3 9

suspended soil (0.444) (1.338) 7

matter

meanlogKd meanlogKd 0.969 -1.903 0.3 9

soil suspended (1.136) (5.703) 7

matter

TheregressionequationswerealsousedtoestimatemeanKdvaluesforsuspendedmatterand

sedimentsfromanestimateofthemeanKd

insoilobtainedfromgeochemicalspeciation

modelingasdiscussedinthenextsection.

3.1.2 EstimationFromGeochemicalSpeciationModeling

Geochemicalspeciationmodelingwasusedtoestimatesoil/waterpartitioningifdata-based

regressionequationscouldnotbeused. ThepartitioningofmetalcationsbetweenDOCand

3-3


20/93

theinorganicportionofthesolutionphasewasalsoestimatedbyspeciationmodeling.Inboth

cases,theU.S.EPAgeochemicalspeciationmodelMINTEQA2,version4.0(Allisonetal.,

1990),wasusedtoestimatetheKdvalues. TheinputdataforMINTEQA2weredeveloped

fromvarioussourcesaspresentedinthefollowingsections.

MODELINGDETAILSANDINPUTDATAFORSOILPARTITIONCOEFFICIENTS

Theconcentrationsofmajorionsusedingeochemicalspeciationmodelingforsoilswerethe

averageconcentrationsinriverwaterasreportedbyStummandMorgan(1996).Thesoil-

waterphosphateconcentrationwasobtainedfromBohnetal.(1979).Theionicstrengthwas

heldconstantat0.005Mafterasensitivitytestintherange0.01to0.001Mrevealedthatthe

impactonresultsofdoingsowassignificantlylessthantheeffectofvariabilityinother

importantparameters. Modelinputvaluesforseveralofthemostsignificantmaster

variablesaffectingKdwerevariedoverreasonablerangesinordertocapturetheexpectedrange

ofKdvalues. ThesemastervariablesincludepH,concentrationofdissolvedorganiccarbon

(DOC),concentrationofparticulateorganiccarbon(POC),andconcentrationofmetaloxide

bindingsites.Therangeforeachofthesemastervariableswascharacterizedbylow,medium,andhighassignedvalues,andthemodelwasexecutedatallpossiblecombinationsofthese

settings. ThepHrangecorrespondedtothatreportedfromtheSTORETdatabase(U.S.EPA,

1996a)withaslightdownwardadjustment(6.5forthemediumvalueinsteadof6.8,and4.5for

thelowvalueinsteadof4.9)toaccountforthemoreacidicenvironmentofsurfacewatershed

soils. TheconcentrationsusedforDOCwere0.5,5.0,and50.0mg/L,takenasareasonable

rangeinsoil-water. TheassignedPOCconcentrationvalueswereobtainedfromanalysisof

datainadatabaseforshallow,silt-loamsoils(Carseletal.,1988andR.Parrish,personal

communication). Thelow,medium,andhighvaluescorrespondedtothe10th,50th,and90th

percentiles,respectively,forparticulateorganicmatterconcentration(0.41,1.07,and2.12

wt%).

Thedominantmetaloxidesorbingsurfacewasassumedtobehydrousferricoxide(HFO).

Becausewehadlittlereliableinformationastotheappropriateconcentrationrange,andalsoin

considerationoftheimportanceofthisvariableindeterminingKd,theHFOconcentrationwas

usedasacalibratingvariable.Thelow,medium,andhighvalueswereinitiallysetto

correspondtothevaluesusedinU.S.EPA(1996a). Thosevalueswerebasedonaspecialized

extractionofreactiveFefromasetof12samplesfromvariousaquifersandsoils.Themean

KdforCd,Cu,Ni,Pb,andZnwerecomputedusingthesevaluesinMINTEQA2. These

computedKdvalueswerecomparedwithmeanKdvaluesforthesesamemetalsinsoilobtained

fromourliteraturesurvey.Thelow,medium,andhighHFOconcentrationswerescaledin

subsequentmodelingsuchthatthemeanKdvaluefromMINTEQA2waswithinthe95%

confidenceintervalofthemeanliteratureKdvalueforeachofthesemetals. (EachMINTEQA2

executionresultedin81differentKdvaluesduetoutilizingalldifferentcombinationsoflow,

medium,andhighassignedvaluesforthefourdifferentmastervariables.Themeanvaluefrom

MINTEQA2wastakenastheaverageofthethreeKdvaluescorrespondingtothemedium

settingofpH,DOC,andHFOandthelowsettingofPOC;themediumsettingsofpH,DOC,

andHFO,andthemediumsettingofPOC;andthemediumsettingsofpH,DOC,andHFO,

3-4


21/93

andthehighsettingofPOC.)AppendixCshowsatypicalMINTEQA2inputfileusedin

estimatingKdforsoil/water.

TheminimumandmaximumKdvalueswereestablishedbycombiningtheavailableliterature

dataandMINTEQA2results.Again,thedistributionwasassumedtobelog-normal. Oncethe

meanlogKd

valueforametalwasestablishedforsoilfromthemodelingexercise,the

previouslydescribedregressionequationsbasedonourliteratureanalysisprocesswereusedto

estimatethemeanKdvaluesforsedimentandsuspendedmatterifthesewerelackingfromthe

literaturedata. Thestandarddeviationwasestimatedasdescribedpreviouslyforthelinear

regressionestimates.

MODELINGDETAILSANDINPUTDATAFOR DOCPARTITIONCOEFFICIENTS

ThepartitioningofmetalsbetweenDOCandotherinorganicformsinwaterisnotusually

reported intermsofapartitioningcoefficient.Infact,specializedalgorithmswithinspeciation

modelsarefrequentlyemployedtoestimatethefractionofmetalboundwithDOCbasedonthe

pH,majorioncompositionofthesolution,andionicstrength. Thedevelopmentofsuch

specializedmethodsforestimatingmetalbindingwithDOCisanongoingresearcharea.MINTEQA2includesaspecializedsub-modelforestimatingDOCinteractionstheGaussian

distributionmodel(Dobbsetal.,1989;AllisonandPerdue,1994). ThismodelrepresentsDOC

asamixtureofmanytypesofmetalbindingsites.Theprobabilityofoccurrenceofabinding

sitewithaparticularlogKisgivenbyanormalprobabilityfunctiondefinedbyameanlogK

andstandarddeviationinlogK. AlimitationoftheDOCbindingcalculationsinMINTEQA2

andsimilarmodelsisthatthemetal-DOCreactionsnecessarytoobtainresultsareknownonly

foralimitednumberofmetalcations,andfornoneoftheanionicmetals.MINTEQA2

includesmeanlogKvaluesforthemetalcationsCd,Cu,Ba,Be,Cr(III),Ni,Pb,andZn.For

othermetalcationsofinterest(Ag,Co,Hg(II),Sn(II),andTl(I)),itwasnecessarytoestimate

themeanlogKforDOCbindingforusewiththeGaussianmodel.ForHg(II),theestimateof

themeanlogKwasdeterminedfromaregressionofknownmeanlogKvaluesagainstthebindingconstantsforhumic-andfulvicacid(HAandFA,respectively)reportedbyTipping

(1994). ThemetalsCd,Cu,Ni,Pb,andZnwerealsorepresentedinthedatabaseofHAandFA

bindingconstants,sothesedatawereusedtodeveloptheregressionrelationshipshownin

Equation(2).

(2)

Asderived,Equation(2)hasacorrelationcoefficient(r2)of0.95,andproducesanestimateof

9.0forthemeanlogKforHg2+bindingwithDOC(meanlogKDOC,Hg). (NOTE: Thisequation

givesthemeanlogKDOC,aformationconstantforuseintheMINTEQA2speciationmodelforthechemicalreactionbetweenDOCandHg.ItisnotthesameasthemeanKdforHgbinding

withDOC. ThelatterisalwaysdesignatedKd;theobjectiveoftheMINTEQA2modelingisto

estimatetheKdforthosemetalsforwhichliteraturedataislacking.)

ThemeanlogKvaluesforDOCbindingtotheothercations(Ag+,Co2+,Sn2+,andTl+)were

derivedfromalinearfreeenergyrelationshipusingthefirsthydrolysisconstants(logKOH)and

3-5


22/93

thebindingconstantforacetate(logKAcet). TheknownvaluesoflogKOHandlogKAcetforthe

metalsCd,Cu,Fe,Ni,Pb,andZnwereusedtoderivethefollowingrelationship:

(3)

Thecorrelationcoefficient(r2)forEquation(3)is0.98. Equation(3)wasusedtoestimatethe

meanlogKDOCvaluesforAg+,Co2+,Sn2+,andTl+foruseinMINTEQA2modeling.Themean

logKDOCvaluesestimatedforthesemetalswere2.0,3.3,6.6,and1.0,respectively.

Theestimationproceduresoutlinedpreviouslyinthissectioncannotreliablybeextendedto

anions. However,anionsaretypicallynotasstronglyboundtoorganicmatter.Therefore,we

usedMINTEQA2toestimateKdvaluesforbindingtoDOCforcationicmetalsonly,and

includedconservativeestimatesofKdvaluesfortheanionsbasedonjudgementalone.

Theconcentrationsofmajorionsusedinestimatingmetal-DOCbindingwithMINTEQA2weretheaverageconcentrationsinriverwaterasreportedbyStummandMorgan(1996).The

concentrationofDOCandthepHweretreatedasmastervariables,witheachassignedthree

levelscorrespondingtolow,medium,andhigh. Theassignedmediumvaluewasthemeanof

thereportedriverandstreamsamplesfromtheliteraturesurvey,andthelowandhighvalues

wereselectedtoencompasstherangeobservedintheliteraturesurveydata.Specifically,the

low,medium,andhighconcentrationsofDOCwere0.89,8.9,and89mg/L,respectively,and

thelow,medium,andhighpHwere4.9,7.3,and8.1,respectively.Thebindingofeachofthe

metalcationswascomputedinninesimulationsthatrepresentedallpossiblecombinationsof

pHandDOCconcentrationlevel. ThemeanKdvalueforeachcationwasspecifiedasthat

valuecomputedbyMINTEQA2whenthepHandDOCconcentrationweresettotheirreported

meanvaluesinsurfacewater. AtypicalMINTEQA2inputfileusedtoestimatemetal

partitioningtoDOCisshowninAppendixD.

TheresultscomputedusingMINTEQA2forbothsoilsandDOCwereusedtoaugmentthe

partitioningdatacollectedintheliteraturesurvey.Althoughitwasconsideredreasonableto

useMINTEQA2toestimatemeanpartitioncoefficients,itwasnotpossibletoestablishthe

shapeoftheKdfrequencydistributioncurvefromtheMINTEQA2results.However,thereis

nocompellingreasontosupposeotherthanthelog-normaldistributionsuggestedbythe

literaturesurveydata.

3.1.3 EstimationfromExpertJudgement

WhenneithertheregressionequationsnorMINTEQA2couldreasonablybeusedtoestimatea

neededmeanlogKd,themeanvaluewasestimatedsubjectivelyusingexpertjudgement.

Factorsconsideredinthisprocessincludedanyvaluesobtainedfromourliteraturesurvey,

reportedmeanvaluesorrangesfrompreviouscompilations,similaritiesofbehavioramong

metals,andqualitativestatementsfromarticlesandreports.TheminimumandmaximumKd

3-6


23/93

valuesfromtheliteraturewerealsousedifreasonablevalueswereavailable.Otherwise,the

extremesinKdwerealsoestimatedbyexpertjudgement.Ineithercase,thestandarddeviation

wasestimatedasdescribedpreviouslyaboveforlinearregressioninSection3.1.1.

Finally,arelativeconfidencelevel(CL)wassubjectivelyassignedtoeachofthefinalvalues

presented. TheCLvaluesrangefrom1to4,withthehighestconfidencecorrespondingtoa

valueof1andthelowesttoavalueof4. Ingeneral,estimatesbasedonourliteraturesurvey

forawell-studiedmetalwithalargeliteraturesamplewasdeemedtomeritaCLof1. Datafor

ametalnotrepresentedintheliteratureforwhichthefinalvalueswerepurelyestimatesfrom

MINTEQA2orothermeanswithanotabledegreeofexpertjudgementinvolvedwereassigned

aCLof4. ManyvaluesweredeterminedincircumstancesthatwarrantedaCLbetweenthese

extremes(e.g.,arangewasgivenintheliterature,avaluewasavailablefromaprevious

compilation,estimatesfromcombinationsoftheselattercircumstancescouldbecombinedwith

estimatesfrommodeling,etc.). Inthesecases,aCLof2or3wasassignedasseemed

appropriate.

Thefinalvaluesweassignedtothemetalpartitioncoefficientsforsoil,sediment,suspendedmatter,andDOCarepresentedinTables3,4,5,and6,respectively.Themethodusedtoarrive

ateachassignedvalue(useofallorasubsetofthecollectedliteratureKdvalues,useof

regressionequations,modelingresults,orexpertjudgement)isindicatedforeachmetaland

media-type,asisthesubjectivelyassignedconfidencelevel.

3-7


24/93

---

---

Table3

Metalpartitioncoefficients(logKd)inL/kgforsoil/soilwater. Valuesinitalicswere

estimatedbyregressionorfromMINTEQA2results.Anentryoflog-normal

indicatesthatthesampledatagaveapositiveresultintheShapiro-Wilktestfor

normalityofthelog-transformeddata.Anentryoflog-normalassumedin

parenthesesmeansthatdatawerenotsufficienttoestablishthedistribution,butlognormalhasbeenassumed. RelativeconfidenceinthedataisindicatedbytheCLvalue

of1to4(1=highest,4=lowest). Anentryof---forthemedianoccurswhere

regressionequationswereusedtoestimatethemean,minimum,andmaximumvalues

andnoestimatewasmadeforthemedian.

Metal Median Mean

Std.

Dev. Min Max Comments

Ag(I) 2.6 2.6 0.8 1.0 4.5 Fromliteraturedata(raw,

n=21);log-normal;CL=1

Asa 3.4 3.2 0.7 0.3 4.3 Fromliteraturedata(raw,n=21);(log-normal

assumed);oxidationstate

usuallynotspecifiedin

literature;CL=2

Ba(II) 2.0 0.7 0.7 3.4 SuspendedmatterK dregressionequationfor

mean;(log-normal

assumed); CL=2

Be(II) 2.2 1.0 1.7 4.1 SuspendedmatterK dregressionequationfor

mean;(log-normal

assumed); CL=3

Cd(II) 2.9 2.7 0.8 0.1 5.0 Fromliteraturedata(edited,


Co(II) 2.1 2.1 1.2 -1.2 4.1 Fromliteraturedata(raw,


Cr(III) 3.9 3.8 0.4 1.0 4.7 Fromliteraturedata(raw,


Cr(VI) 1.1 0.8 0.8 -0.7 3.3 Fromliteraturedata(raw,

n=24);(log-normal

assumed); CL=2

Cu(II) 2.7 2.5 0.6 0.1 3.6 Fromliteraturedata(raw,


3-8


25/93

---

Table3









Metal Median Mean

Std.


Hg(II) 3.8 3.6 0.7 2.2 5.8 Fromliteraturedata(raw,


MeHg 2.8 2.7 0.6 1.3 4.8 Fromliteraturedata(raw,n=11);log-normal;CL=2

Mo(VI) 1.1 1.3 0.6 -0.4 2.7 Fromliteraturedata(raw,

n=5);(log-normal

assumed);oxidationstate

notalwaysspecifiedin

literaturedata;CL=3

Ni(II) 3.1 2.9 0.5 1.0 3.8 Fromliteraturedata(raw,


Pb(II) 4.1 3.7 1.2 0.7 5.0 Fromliteraturedata(edited,n=31);(log-normal

assumed); CL=2

Sbb 2.3 1.1 0.1 2.7 Fromliteraturedata(mean

istheaverageofseveral

reportedmeanvalues,n=5);

(log-normalassumed);

CL=4

Se(IV)c 1.4 1.3 0.4 -0.3 2.4 Fromliteraturedata(edited,

n=11);(log-normal

assumed); CL=2

3-9


26/93

---

---

---

---

---

Table3









Metal Median Mean

Std.


Se(VI) -0.2 1.1 -2.0 2.0 Meanestimatedfrom

MINTEQA2result;(log

normalassumed);min,max

fromexpertjudgement;

CL=4

Sn(II) 2.7 0.7 2.1 4.0 Fromliteraturedata;(log

normalassumed);CL=3

Tl(I) 0.5 0.9 -1.2 1.5 Estimatedfrom

MINTEQA2result;(log

normalassumed);CL=4

V(V) 1.7 1.5 0.5 2.5 Mean,min,maxfrom

suspendedmatterKdregressionequation;(log

normalassumed);CL=4

Zn(II) 3.1 2.7 1.0 -1.0 5.0 Fromliteraturedata(raw,

n=21);(log-normal

assumed); CL=1

CN- 0.7 1.6 -2.4 1.3 Estimatedfrom

MINTEQA2result;(log

normalassumed);CL=4

a PublishedpartitioningdataforAsdoesnotallowdifferentiationofAs(III)andAs(V).Itisprobablethatpublishedvaluesrepresentresultsinvolvingbothoxidationstates.

b PublishedpartitioningdataforSbisrareanddoesnotallowdifferentiationofSb(III)

andSb(V).

PositiveresultinShapiro-Wilktestfornormalityofdatanotlog-transformed.But

samplesizeissmallanddatamaynotbeveryrepresentative.

3-10

c


27/93

---

---

---

---

---

Table4

Metalpartitioncoefficients(logKd)inL/kgforsediment/porewater. Valuesinitalics

wereestimatedbyregressionorfromMINTEQA2results.Anentryoflog-normal







Metal Median Mean

Std.


Ag(I) 3.6 1.1 2.1 5.8 MeanfromsoilK dregressionequation;(log

normalassumed); min,max

fromliteraturedata;CL=3

Asa 2.2 2.4 0.7 1.6 4.3 Fromliteraturedata;log

normal;oxidationstatenot

specifiedinliteraturedata;

CL=2

Ba(II) 2.5 0.8 0.9 3.2 Mean,min,maxfrom


normalassumed);CL=3

Be(II) 2.8 1.9 0.8 6.5 Mean,min,maxfrom


normalassumed);CL=3

Cd(II) 3.7 3.3 1.8 0.5 7.3 Fromliteraturedata(n=14,

edited);log-normal;CL=1

Co(II) 3.1 1.0 2.9 3.6 MeanfromsoilK dregressionequation;(log



Cr(III) 4.9 1.5 1.9 5.9 Mean,min,maxfromsoil

Kdregressionequation;


CL=4

3-11


28/93

---

---

---

---

---

Table4









Metal Median Mean

Std.


Cr(VI) 1.7 1.4 0.0 4.4 Mean,min,maxfromsoil



CL=4

Cu(II) 4.1 3.5 1.7 0.7 6.2 Fromliteraturedata(raw,n

=12);log-normal; CL=1

Hg(II) 4.9 0.6 3.8 6.0 Fromliteraturedata(raw,

n=2);(log-normal

assumed); CL=2

MeHg 3.9 0.5 2.8 5.0 Fromliteraturedata(edited,

n=2);(log-normal

assumed); CL=2

Mo(VI) 2.5 0.8 0.4 3.7 Meanfromliteraturedata

(reportedmeanvaluewith

oxidationstatenot

specified);(log-normal

assumed);min,maxfrom

soilKdregressionequation;

CL=4

Ni(II) 3.9 1.8 0.3 4.0 MeanfromsoilK dregressionequation;(log

normalassumed); min,maxfromliteraturedata;CL=3

Pb(II) 5.1 4.6 1.9 2.0 7.0 Fromliteraturedata(edited,


3-12


29/93

---

---

---

---

---

---

Table4









Metal Median Mean

Std.


Sbb 3.6 1.8 0.6 4.8 Fromliteraturedata

(reportedmeanvalue);(log

normalassumed);CL=4

Se(IV) 3.6 1.2 1.0 4.0 Meanfromliteraturedata

(reportedmeanvalue);(log


fromexpertjudgement;

CL=4

Se(VI) 0.6 1.2 -1.4 3.0 Mean,min,maxfromsoil



CL=4

Sn(II) 3.7 0.7 3.1 5.1 Mean,min,maxfromsoil



CL=3

Tl(I) 1.3 1.1 -0.5 3.5 Mean,minfromsoilKdregressionequation;(log

normalassumed); max


V(V) 2.1 0.9 0.4 3.2 Mean,min,maxfrom


normalassumed);CL=4

Zn(II) 4.8 4.1 1.6 1.5 6.2 Fromliteraturedata(edited,

n=13);(log-normal

assumed); CL=1

3-13


30/93

---

Table4









Metal Median Mean

Std.


CN- 1.6 1.7 -1.8 2.2 Mean,min,maxfromsoil



CL=4

a PublishedmetalpartitioningdatadoesnotallowdifferentiationofAs(III)andAs(V).Itis

probablethatthedatapresentedincluderesultsforbothoxidationstates.b PublishedpartitioningdataforSbisrareanddoesnotallowdifferentiationofSb(III)and

Sb(V).

3-14


31/93

---

Table5

Metalpartitioncoefficients(logKd)inL/kgforsuspendedmatter/water. Valuesinitalicswere

estimatedbyregressionorfromMINTEQA2results.Anentryof log-normalindicatesthat

thesampledatagaveapositiveresultintheShapiro-Wilktestfornormalityofthelog-

transformeddata.Anentryoflog-normalinparenthesesmeansthatdatawerenotsufficient

toestablishthedistribution,butlog-normalhasbeenassumed. RelativeconfidenceinthedataisindicatedbytheCLvalueof1to4(1=highest,4=lowest). Anentryof---forthemedian

occurswhereregressionequationswereusedtoestimatethemean,minimum,andmaximum

valuesandnoestimatewasmadeforthemedian.

Metal Median Mean

Std.


Ag(I) 5.2 5.2 0.6 4.4 6.3 Fromliteraturedata(edited,n=9);

log-normal;CL=2

Asa 4.0 3.9 0.5 2.0 6.0 Fromliteraturedata(raw,n=25);

(log-normalassumed);oxidationstatenotspecifiedintheliterature

data;CL=2

Ba(II) 4.0 4.0 0.4 2.9 4.5 Fromliteraturedata(raw,n=14);

log-normal;CL=2

Be(II) 4.1 4.2 0.7 2.8 6.8 Fromliteraturedata(raw,n=17);

log-normal;CL=2

Cd(II) 5.0 4.9 0.6 2.8 6.3 Fromliteraturedata(edited,


Co(II) 4.7 4.8 0.8 3.2 6.3 Fromliteraturedata(edited,


Cr(III) 5.1 5.1 0.4 3.9 6.0 Fromliteraturedata(raw,n=25);

log-normal;assumesunspecified

oxidationstateis(III);CL=2

Cr(VI) 4.2 0.5 3.6 5.1 Mean,min,maxfromsoilKdregressionequation;(log-normal

assumed); CL=4

Cu(II) 4.7 4.7 0.4 3.1 6.1 Fromliteraturedata(edited,n=42);log-normal;CL=1

Hg(II)b 5.3 5.3 0.4 4.2 6.9 Fromliteraturedata(edited,


3-15


32/93

---

---

---

---

---

---

Table5








Metal Median Mean

Std.


MeHg 4.9 0.7 4.2 6.2 MeanfromsoilK dregression

equation;(log-normalassumed);

min,maxfromliteraturedata;

CL=3

Ni(II)b 4.3 4.4 0.4 3.5 5.7 Fromliteraturedata(edited,


Mo(VI) 4.4 1.0 3.7 4.9 Mean,min,maxfromsoilKdregressionequation;(log-normal

assumed); CL=4

Pb(II)c 5.7 5.7 0.4 3.4 6.5 Fromliteraturedata(edited,

n=38);(log-normalassumed);

CL=1

Sbd 4.8 0.5 3.9 4.9 Mean,min,maxfromsoilKdregressionequation;(log-normal

assumed); CL=4

Se(IV) 4.4 0.4 3.8 4.8 Mean,min,maxfromsoilKdregressionequation;(log-normal

assumed); CL=4

Se(VI) 3.8 1.0 3.1 4.6 Mean,min,maxfromsoilKdregressionequation;(log-normal

assumed); CL=4

Sn(II) 4.9 0.8 4.7 6.3 Mean,minfromsoilKdregressionequation;(log-normalassumed);

maxfromliteraturedata;CL=4

3-16


33/93

---

---

---

c

Table5








Metal Median Mean

Std.


Tl(I) 4.1 1.0 3.0 4.5 MeanfromsoilKdregression

equation;(log-normalassumed);

otherparametersfromexpert

judgement;CL=4

V(V) 3.7 0.6 2.5 4.5 Meanfromliteraturedata(raw,

n=5);(log-normalassumed);min,

maxfromexpertjudgement;

oxidationstatenotalways

specifiedinliterature;CL=3

Zn(II) 5.1 5.0 0.5 3.5 6.9 Fromliteraturedata(edited,


CN- 4.2 0.6 3.0 4.4 Mean,min,maxfromsoilKdregressionequation;(log-normal

assumed); CL=4

a PositiveresultforShapiro-Wilktestfornormalityofdatanotlog-transformed.Published

metalpartitioningdatadoesnotallowdifferentiationofAs(III)andAs(V).Itisprobable

thatthedatarepresentedincluderesultsforbothoxidationstates.b FailedShapiro-Wilktestfornormalityoflog-transformeddata,butpassedthe

Kolmogorov-Smirnovtestandhistogramexhibitslog-normalcharacter.

FailedShapiro-WilkandtheKolmogorov-Smirnovtestfornormalityoflog-transformed

data,buthistogramexhibitslog-normalcharacterd PublishedpartitioningdataforSbisrareanddoesnotallowdifferentiationofSb(III)and

Sb(V).

3-17


34/93

Table6

Metalpartitioncoefficients(logKd)inL/kgforpartitioningbetweenDOCand

inorganicsolutionspecies. Valuesinitalicswereestimatedbyregressionorfrom

MINTEQA2results.Log-normaldistributionsareassumed.Relativeconfidenceinthe

dataisindicatedbytheCLvalueof1to4(1=highest,4=lowest).

Metal Mean

Std.

Dev. Min Max Comment

Ag(I) 2.5 1.0 1.5 4.5 MeanestimatedfromMINTEQA2

results;otherparametersfromexpert

judgement;(log-normalassumed);

CL=3

As 2.0 1.0 0.0 3.0 Nodata,valuesfromexpertjudgement

(conservative);(log-normalassumed);

(log-normalassumed);CL=4

Ba(II) 3.6 1.0 2.5 4.0 MeanestimatedfromMINTEQA2

results,valuesforotherparametersfrom

expertjudgement;(log-normal

assumed); CL=3

Be(II) 2.1 1.0 1.1 3.8 Allparametersestimatedfrom

MINTEQA2results;CL=3

Cd(II) 3.8 0.9 2.0 5.5 MeanestimatedfromMINTEQA2

results;min,maxfromexpert

judgement;CL=3

Co(II) 3.8 0.9 2.0 5.5 MeanestimatedfromMINTEQA2


judgement;CL=3

Cr(III) 1.1 1.6 -0.6 4.3 MeanestimatedfromMINTEQA2


judgement;CL=4

Cr(VI) 2.0 1.0 0.0 3.0 Nodata,valuesfromexpertjudgement


CL=4

Cu(II) 5.4 1.1 2.5 7.0 Fromliteraturedata(raw,n=17);(log

normalassumed);CL=2

Hg(II) 5.4 1.2 3.0 6.0 Meanfromliteraturedata(raw,n=3);

(log-normalassumed); min,maxfrom

expertjudgement;CL=4

3-18


35/93

Table6





Metal Mean

Std.


MeHg 5.0 1.1 2.8 5.5 Mean,min,maxestimatedbasedon

relativeKdsofHg(II)andMeHgfor

suspendedmatterandHg(II)Kdwith

DOC; (log-normalassumed);CL=4

Ni(II) 3.7 0.9 1.9 5.4 MeanestimatedfromMINTEQA2



CL=3

Mo(VI) 2.0 1.0 0.0 3.0 Nodata,valuesfromexpertjudgement


CL=4

Pb(II) 4.9 0.5 3.8 5.6 Fromliteraturedata(raw,n=9);(log

normalassumed);CL=2

Sb 2.0 1.0 0.0 3.0 Nodata,valuesfromexpertjudgement


CL=4

Se(IV) 2.0 1.0 0.0 3.0 Nodata,valuesfromexpertjudgement


CL=4

Se(VI) 2.0 1.0 0.0 3.0 Nodata,valuesfromexpertjudgement


CL=4

Sn(II) 2.0 1.0 0.0 3.0 Nodata,valuesfromexpertjudgement


CL=4

Tl(I) 1.6 1.0 0.0 3.0 MeanestimatedfromMINTEQA2,

valuesforotherparametersfromexpert


CL=4

3-19


36/93

Table6





Metal Mean

Std.


V(V) 2.0 1.0 0.0 3.0 Nodata,valuesfromexpertjudgement


CL=4

Zn(II) 5.1 0.7 4.6 6.4 Fromliteraturedata(raw,n=9);(log

normalassumed);CL=3

CN 2.0 1.0 0.0 3.0 Nodata,valuesfromexpertjudgement

(conservative);(log-normalassumed);CL=4

3-20


37/93

3.2DEVELOPMENTOFPARTITIONINGCOEFFICIENTSFORWASTESYSTEMS

Themultimedia,multi-pathwayriskassessmentfor3MRAutilizesasourcemodelthatassumes

equilibriumpartitioninginlandapplicationunits(LAUs),wastepiles,landfills,treatmentlagoons

(surfaceimpoundments),andaeratedtanks. Theavailabledataforcharacterizingthepartitioningof

metalsinwasteconsistsalmostexclusivelyofleachateextractiontestresultsforspecificwastes.

Ourliteraturesearchdidnotproduceanystudythatspecificallyprovidesmeasuredpartitioning

coefficientsformetalsinthemixedmaterialspresentinwastemanagementunits.

Severalstudieshaveaddressedtheissueoftheapplicabilityofleachateextractiontestdatato

predicttheleachatecompositionexitinglandfills(U.S.EPA,1991).TheU.S.EPAToxicity

CharacteristicLeachingProcedure(TCLP)wasspecificallydesignedtocharacterizeleachate

compositionsproducedbyspecificwastesco-disposedwithmunicipalsolidwaste.Recentpapers

suggestthattheconcentrationobservedinanyleachtestdependsagreatdealonleachingtimeand

thecumulativesolid-liquidratio(vanderSlootetal.,1996). Threeregimesarerecognizedinthe

leachingprocess(deGrootandvanderSloot,1992).Inthefirstregime,theleachatecompositionis

controlledbyinitialwash-offoflooselyadheredcontaminant;inthesecond,theleachatecompositioniscontrolledbydissolutionofprimarymaterialsandperhapsre-precipitationofmore

stablephases;andinthethird,theleachatecompositioniscontrolledbythediffusionofwaste

constituentsfromtheinteriorofwasteparticlestotheparticlesurface.Thetimeofonsetand

durationoftheseregimesarehighlyvariable,anddependonthelife-cycleofthespecificwaste

system(acetogenesis,methanogenesis,etc.).Theoverallchemicalcomposition(majorion

concentrationandconcentrationofmetal-complexingorganicligands)isalsoimportantin

determiningthemetalleachateconcentrationthatwillbeobservedinanyparticularcase.In

general,itwouldseemthatthehighestmetalleachateconcentrationswouldbeexpectedduringthe

initialwash-offperiod,withconcentrationsdecliningthereafter.Animmediatelyobviousquestion

is:Whatperiodisofconcerninthemodelingforthe3MRAasusedforHWIRrulemaking?Since

the3MRAmodeldoesnotallowatime-variablepartitioncoefficient,itwouldseemthatanaggregatepartitioncoefficientthatrepresentsanaverageoveranappropriateexposureorwaste

managementunitlifetimewouldbedesired. Unfortunately,thereiscurrentlynowaytoknow

whetherthepartitioningobservedinaTCLPtestcorrespondstosuchanaveragevalue. Most

authorsseemtoregardtheTCLPasanaggressivetestthatmayoverestimatemetalleachate

concentrations. However,thereisnoconsensusonthispoint.

Inviewofthelackofdatadescribingpartitioningofmetalsindifferenttypesofwasteunits,the

followingsimplificationsareproposed:

1) Forlandapplicationunits,thepartitioncoefficientsforsoilspresentedinTable3shouldbe

used. ThissimplificationassumesthatthepartitioningbehaviorofmetalsinanLAUislikelytobedominatedbythesorptivecharacteristicsofthesoilunderlyingtheunit.

2) Forsurfaceimpoundmentsandaeratedtanks,thepartitioncoefficientsforsuspendedmatter

presentedinTable5shouldbeused. Thisseemsareasonablestepinthatpartitioninginsuch

systemsmustinvolvesorptiontosuspendedparticlesandsediments. Thecompositionand

3-21


38/93

quantityofsuspendedandsedimentedsorbingparticlesmustbequitevariable,butthereisno

sourceofdataonwhichtobasemorespecificmodelingorotherestimatingtechniques.

3) Wastepilesandlandfillsshouldbetreatedthesameasregardsmetalpartitioning.

Adoptingthesesimplifications, itisonlynecessarytoderiveseparateestimatesofmetalpartition

coefficientsspecificallyforwastepilesandlandfills.Thefollowingsectionsdetailhowtheselatter

twosetsofwastemanagementunitcoefficientshavebeenestimatedfromavailableTCLPand

similarleachateextractionteststhatcharacterizeboththesolidphaseandthecorresponding

leachatemetalconcentrations. Wehavealsousedstatisticalmethodsandgeochemicalspeciation

modelingtoextendresultstometalsnotrepresentedinreportedTCLPorotherleachtestresults,

andtoexaminethesimilaritybetweenexpectedwastepartitioningandpartitioninginnaturalmedia.

3.2.1 EstimationfromAnalysisofDataPresentedintheLiterature

TherearenumerouspapersandjournalarticlesdescribingresultsfromaTCLPorsimilarleachtest

foraparticularwaste. Thesepublishedstudiesoftenfocusonwasteconstituentleachabilitybeforeandafterawastestabilizationortreatmentprocess. Therearemanypublishedstudiesofthe

leachabilityofmetalsfromincineratorash,withtheaimofinvestigatingthesuitabilityoftheash

materialsfordisposalorforuseinconstruction.Unfortunately,leachateextractiontestresults

(metalleachateconcentrations)oftenarereportedwithoutthecorrespondingconcentrationinthe

solidphase. Thisomissionmakesthosedatauselessinestimatingexpectedmetalspartitioning.

Ourliteraturesurveyproduced203leachtestresultsforwhichbothleachateandsolidphasedata

werepresented.Table7showstherangeandmeanvaluesofeffectivepartitioncoefficients

calculatedforeachmetalforwhichsufficientdatawasfound.Werefertotheseaseffective

partitioncoefficientsbecausetheyaresimplytheratioofmetalconcentrationinthesolidphaseto

thatinthesolutionphaseasrepresentedintheleachtestresults. Thesecoefficientsmayormaynot

representequilibriumpartitioning.

Severalauthorsdiscussedthesimilaritiesinmetalleachabilityoverarangeofdifferentmaterials.A

studybyvanderSlootetal.(1996)examinedtheleachingbehaviorofCdandZnfromvariousash

materials,shreddedmunicipalsolidwaste,sewagesludge-amendedsoil,andsoil.Similar

characteristicswerenotedinpHdependentleachingofbothCdandZnfromtheninedifferent

materialsstudied.Differencesamongthedifferentmaterialswereattributedtowaste-specific

chemicalparametersthatcausedadifferentchemicalspeciation.Forexample,theauthorscite

possibleCdcomplexationwithchloridethattheyinvestigatedusingMINTEQA2.Theyfoundthat

anincreasedleachabilityofCdinsomeoftheashmaterialswascorrelatedwithincreasedchloride

concentrationinthewaste.

Flyhammar(1997)concludedthattherearesimilaritiesinthemetalbindingpropertiesofmunicipal

solidwaste(MSW)andsediments. Hefoundthatthefractionationofmetalsamongvarious

availableandreactiveforms(asdeterminedbysequentialchemicalextractions)wassimilarbetween

freshMSWandanoxicsediment. Similaritieswerealsofoundinthefractionationpatternsofaged

MSWandanoxicsediments.

3-22


39/93

Table7

Effectivemetalspartitioncoefficientsbasedonreportedsolidphaseandsolutionphasemetals

concentrationsfromleachtestsreportedintheliterature.Nisnumberofsamples;meanand

rangeareexpressedinlogunits(L/kg).

Metal N Mean Range

As 11 2.8 1.0-5.1

Ba 7 3.0 1.8-3.7

Be 2 2.8 2.7-6.8

Cd 31 1.3 0-3.9

Co 6 2.8 1.6-3.8

Cr(III) 27 3.0 0.6-6.2

Cr(VI) 6 4.1 2.2-6.2

Cu 16 3.3 2.0-5.1

Hg 8 3.1 1.7-4.4

Ni 12 2.3 1.3-4.7

Pb 31 2.7 0.0-4.9

Sb 4 2.7 1.7-3.2

V 4 2.9 2.7-3.1

Zn 23 2.6 1.2-4.7

TheconsistencyinthemetalspartitioningaffinityrelationshipsnotedinSection3.1.1andthe

similaritiesnotedbytheselatterauthorsinthefractionationandbehaviorofmetalsinwasteversus

thatinsoilsandsedimentsleadstothesuppositionthatthepartitioningbehaviorofmetalsinmixed

wastesystemsmightnotbealtogetherdifferentfromthatinanaturalmedium.Itwouldperhapsbe

surprisingiftherelativeaffinitiesfordifferentmetalsinwasteweremarkedlydifferentfromtheir

relativeaffinitiesinnaturalmaterials. Theremaycertainlybesomedeviationsduetothepresence

ofoneormorecomplexingagentsinwastesystemsthathaveapreferenceforcombiningwith

certainofthemetals;however,intheabsenceofdatatoquantifythiseffect,andalsoin

considerationofthepaucityofactualpartitioningdataforwastesystems,wehavedevelopeda

regressionequationthatpredictswasteKdfromsoilKdforusewithmetalsforwhichlittleorno

datawasfound.WechosetousesoilKdasthepredictorbecauseacomparisonofKdvaluesfor

soils,sedimentsandsuspendedmattersuggestedthatthesolidtoliquidconcentrationratiois

importantindeterminingthemagnitudeofKd. (ThisapparentdependenceofKdonsolidtoliquid

ratiohasbeennotedinotherstudiesandissometimesreferredtoastheparticleconcentration

3-23


40/93

effect.) Thissolidtoliquidconcentrationratioforlandfillsandwastepilesisprobablymore

similartothatofsoilsthantoanyothermedium.Also,wenotethatlandfilledwasteistypically

coveredwithsoiltoformsoil/wastelayerswithinalandfillcell.Indevelopingourregression

relationship,weusedtheeffectivepartitioncoefficientsforthemetalsforwhichwehadthemost

complete(largest)sample.Theregressionequationthusdeterminedis:

(4)

Thisrelationshiphasaratherlowcorrelationcoefficient(r2)of0.4implyingthatonly40%ofthe

variationinlogKd,wastefromtheleachtestdataisaccountedfor.TheimplicationisthatlogKdvaluespredictedbymeansofthisequationmustberegardedashighlyuncertain.AppendixB

showsthescatterplotofdatafromwhichthisrelationshipwasdeveloped.

3.2.2 EstimationfromGeochemicalSpeciationModeling

TheMINTEQA2geochemicalspeciationmodelwasusedtoinvestigatethepossiblerangeofmetal

partitioncoefficientsforlandfills. Theinputrequirementsofthemodelforestimatingmetalpartitioningincludetheconcentrationsofmajorsoluteions,thepH,theconcentrationsofsorbing

phases,andtheDOCconcentration. Fourlandfillmodelingscenariosweredeveloped,

distinguishedprimarilybytheconcentrationsofmajorsoluteions,theDOCconcentration,thePOC

concentration,andthepH.Thesescenariosincludedlandfillscontainingmunicipalsolidwastein

theacetogenicstageandinthemethanogenicstage,amonofillcontainingashfromincinerationof

municipalsolidwaste(MSWIash),andamonofillcontainingcementkilndust(CKD).

ForeachoftheMINTEQA2modelingscenarios,ahydrousferricoxidesorbingphasewasassumed.

AparticulateorganiccarbonsorbentwasalsoassumedfortheacetogenicandmethanogenicMSW

landfills.Particulateorganiccarbonwasassumedtohavebeenconsumedintheincinerationprocess

fortheMSWIandCKDscenarios. Theconcentrationofthesorbentiscrucialindeterminingthe

numberofsitesavailableformetalsorption. Unfortunately,theconcentrationofsorbent

appropriateinthevariouswastemanagementsystemsissubjecttoaveryhighdegreeofuncertainty.

Theuncertaintyarisesfromthevariablecompositionofwastesthataredisposedinlandfillsandthe

possiblechangesincompositionovertimeasleachatepercolatesthroughthematerials.Itislikely

thatsolidsurfacesexposedtolandfillgasandleachateundergochangeswithrespecttotheir

sorptivecharacterovertime. Possiblechangesincludedissolutionorprecipitationofoxideor

organicsurfacecoatings.Theseprocesseshavenotbeenstudiedinactuallandfillsamplesin

sufficientdetailtoallowquantitativerepresentation. Kerstenetal.(1997)citedevidenceofsorption

controlofPbleachinginMSWIleachtests.TheyattemptedtomodeltheobservedPb

concentrationsbyutilizingaspeciationmodelwithsurfacecomplexationsorptionreactionsparameterizedfortheconstantcapacitancemodelassuminghydrousferricoxide(HFO)asthe

sorbent.Theyobtainedreasonableresultsassuming0.7g/LfortheHFOconcentrationandusinga

sitedensityof1.35x10-4molsites/gHFO. TheMINTEQA2modelingpresentedhereutilizeda

similarsurfacecomplexationmodel(thediffuse-layermodel).Kerstenetal.(1997)hadnotedthat

theirsorbentconcentrationwasperhapstoolow,soourmodelingwasconductedbothwiththeir

3-24


41/93

valueof0.7g/L,andusing7g/Lasareasonableupper-rangevalue.Inbothcases,asitedensity

1.35x10-4molsites/gHFOwasused.

Thevaluesofotherparametersandconstituentconcentrationsusedinourmodelingforthefour

landfillscenariosareshowninTable8. Afterconcentrationofsorbingsites,themostcriticalmodel

parameterispH,sothemodelingwasconductedatthreedifferentpHvaluesforeachscenario.The

threepHvaluesusedfortheacetogenicandmethanogenicscenarios(4.5,6.1,7.5and7.5,8.0,9.0,

respectively)wereinkeepingwiththeminimum,maximumandmeanpHcitedfortheselandfill

stagesinastudyof15landfillsbyEhrig(1992).Themajorionconcentrationsfortheacetogenic

andmethanogenicscenarioswerealsoasspecifiedinEhrig(1992).ThethreepHvaluesforthe

MSWIscenario(8.0,9.0,10.0)wereselectedtodefineareasonablerangeandcentraltendency

valueforthisscenario. Thesevalueswerebasedondatacollectedintheliteraturereviewportionof

thisstudy,aswerethemajorionconcentrationsfortheMSWIscenario.ThepHvaluesassociated

withtheCKDscenario(9.0,10.0,11.0)wereselectedwithdueconsiderationtothehighlyalkaline

conditionsassociatedwiththismaterial,buttheylackstatisticalsignificance.Anexample

MINTEQA2inputfileforeachofthescenariosispresentedinAppendixE.

Itshouldbenotedthattheconfidencelevelassociatedwithallofthemodelingparametersforwaste

systemsislow.Thereisnotanextensivedatabaseofobservationsfromwhichtoextractreasonable

modelvaluesformostoftheseparameters,especiallytheconcentrationofsorbentsandsorbing

sites. Withoutreliableinformationforcharacterizingthesorbents,itisnotpossibletoaccurately

establishthetotalsystemconcentrationsofcompetingions(Ca,Mg,etc.)thatshouldbeusedinthe

model. Theresultsmustbeinterpretedinlightofthisshortcoming.

Table8

ImportantparametersandconstituentconcentrationsusedinMINTEQA2

modelingoflandfillsintheacetogenicandmethanogenicstagesandMSWIand

CKDmonofills.

Model

Parameter

Scenario

MSW

Acetogenic

MSW

Methanogenic

MSWI

Ash

Monofill

CKD

Monofill

pH 4.5,6.1,7.5a 7.5,8.0,9.0a 8.0,9.0,

10.0b9.0,10.0,

11.0c

IonicStrength(M)

0.1c 0.1c 0.1c 0.1c

Ca (mg/L) 6000d 975d 1,700b 2850f

Mg

(mg/L)

625d 500d 10b 10f

3-25


42/93

c

Table8

ImportantparametersandconstituentconcentrationsusedinMINTEQA2

modelingoflandfillsintheacetogenicandmethanogenicstagesandMSWIand

CKDmonofills.

Model

Parameter

Scenario

MSW

Acetogenic

MSW

Methanogenic

MSWI

Ash

Monofill

CKD

Monofill

Na (mg/L) 1350e 1350e 300b 300f

K (mg/L) 1100e 1100e 380b 400f

CO3 (mg/L) 500c 250c 50c 50f

Cl (mg/L) 2100e 2100e 1,200b 380f

Fe (mg/L) 780e 0 0 0

SO4 (mg/L) 500d 80d 1,400b 630f

DOC (mg/L) 100c 50c 15c 15c

POC (mg/L) 100,000c 50,000c 0 0a Minimum,average,andmaximumvaluesreportedinEhrig(1992).

b ObtainedfromanalysisofMSWIdataobtainedinourliteraturesurvey.

Reasonableguesses.d ComputedfromtypicaldissolvedvaluesreportedinEhrig(1992),assumingequilibrium

withthemodelsorbentsatthemedianpHforacetogenicandmethanogeniccases.e ReportedastypicalvaluesinEhrig(1992).f GeneratedfromsimulationofTCLPonCKDusingMINTEQA2(U.S.EPA,1998b).

ThepartitioningcoefficientsforselectedwastesestimatedfromtheMINTEQA2modelingexercise

forseveralmetalsareshowninTable9. Thepartitioncoefficientswerecalculatedastheratioof

thesimulatedsorbedanddissolvedconcentrationsasexpressedinEquation(1). TheunitsofKdwereconvertedtoL/kgbyassumingthatoneliterofleachatesolutionisassociatedwith5kgof

wastematerial. Therangeinestimatedpartitioncoefficientsisshownforeachlandfillmodeling

scenario. Ininterpretingtheseresults,itmustberememberedthatnostatisticalsignificancecanbe

assignedbecausenonecanbeassociatedwithmostofthemodelinputparameters.Atbest,these

resultsshouldberegardedasindicatingapossiblerangeofcentraltendencyvalues,andeventhismustbequalifiedbecausetheresultsaresosensitivetoseveralpoorlycharacterizedparameters,

mostnotably,theconcentrationofsorbents.Theresultsalsoreflectonlyasinglesetof

concentrationvaluesforthemajorambientionsvariabilityintheseconcentrationswillinfluence

metalpartitioning. Someionsexertgreaterinfluenceonthepartitioningofparticularmetals.For

example,thelowpartitioncoefficientsassociatedwithCdinTable9appeartoberelatedto

complexationwithchloridethatisenteredatrelativelyhighambientconcentrationinallscenarios.

3-26


43/93

Thiseffectisinkeepingwithobservationsbyothers(vanderSlootetal.,1996).Anothermajor

ambientionwhoseconcentrationlevelcaninfluencemetalpartitioningiscalcium.Atthehigh

concentrationsofcalciumcationsfoundinwastesystems,especiallyMSWIashandCKD,the

competitionforbindingsitescanbecomeveryimportantwithregardtotracemetalbinding.For

thosetracemetalswhosepartitioningissignificantlyinfluencedbytheconcentrationlevelofa

majorambientionsuchaschlorideorcalcium,itisexpectedthatthisfactalonewouldcontributeto

abroaderrangeofobservedpartitioncoefficientsinrealsystemsthanthatcalculatedinthis

modelingexercise.

Table9

Estimatedrangeinlogpartitioncoefficients(L/kg)inwasteforselectedmetalsdetermined

fromMINTEQA2modeling.

Metal

EstimatedlogKd(L/kg)

MSWAcetogenesis

MSWMethanogenesis

MSWI

AshMonofill

CKDMonofill

Be 0.8-3.9 3.3-4.4 (-0.4)-4.0 (-2.7)-2.4

Cd (-0.3)-0.0 0.6-1.7 (-1.0)-1.1 (-0.4)-1.2

Co 0.2-0.3 0.9-1.8 (-0.9)-0.4 (-2.0)-0.2

Cr(III) 1.1-3.5 3.8-4.8 (-0.2)-3.2 (-2.5)-2.3

Cu 1.1-1.9 2.0-2.5 0.0-2.9 (-2.0)-2.1

Ni 0.2-0.4 1.1-1.9 (-0.04)-1.1 (-1.5)-0.9

Pb 1.7-2.7 3.3-4.2 2.4-3.6 0.7-3.4

Zn 0.4-0.7 1.5-2.1 (-0.6)-1.3 (-2.7)-1.1

WecomparedthepartitioncoefficientsestimatedforwastesusingMINTEQA2withvalues

predictedbythepreviouslydiscussedregressionequation(logKd,waste=0.7logKd,soil+0.3;see

Section3.2.1). Thedegreeofagreementvariedamongmetals.(Wedefinedthemeasureof

agreementforametaltobewhetherthevaluepredictedbytheregressionequationusingthemean

soilKdvalueofTable3fallswithintherangeofMINTEQA2estimatesforthatmetal.Usingthis

ratherlaxrequirementforagreement,theMINTEQA2-modeledKdvaluesforBe,Cr(III),Cu,and

Pbagree,thoseofCdandNidonotagree,andthoseofCoandZnaremarginal.)In

agreementwiththeliterature-reportedKdvaluesfornaturalmedia,PbandCr(III)tendtohavehigh

KdestimatesfromtheMINTEQA2wastesimulations.Ingeneral,theMINTEQA2resultsforthe

acetogenicandmethanogeniclandfillscenariosagreedmorecloselywithvaluesestimatedbythe

regressionrelationshipbasedonsoilKdvaluesthanforthemorealkalineashandCKDlandfill

3-27


44/93

scenarios. ItisprobablethatthelowerKdvaluesinthelatterscenariosareduetothecombination

ofhighermajorambientionconcentrationsthatcompetewiththetracemetalsforsorbingsitesand

solubilizethemetalsbycomplexation,plustheassumedabsenceofparticulateorganiccarboninthe

modellandfillsystems.

3-28


45/93

4.0DISCUSSIONOFRESULTSANDSOURCESOFUNCERTAINTY

Partitioncoefficientsobtainedfromliteraturedataaresubjecttonumeroussourcesofuncertainty.

Manypreviousstudieshavedemonstratedthatinavarietyofsoilsandforavarietyofmetals,

partitioncoefficientsvarywithpHandwiththeconcentrationofsorbingphasesinthesoilmatrix

(e.g.,weightpercentorganicmattercontentandweightpercenthydrousferricoxidesand

correspondingoxidesofaluminumandmanganese)(Janssenetal.,1997;HassanandGarrison,

1996;BangashandHanif,1992;AndersonandChristensen,1988).Itiswellknownthatdissolved

ligandspresentinsoilporewater(e.g.,dissolvedorganicmatter,anthropogenicorganicacids)can

complexwithmetals,reducingtheirpropensityforsorptioninproportiontotheconcentrationofthe

ligands(Christensenetal.,1996). Inmulti-metalsystems,competitionamongmetalsforsorption

sitesandtheattendantreductioninthepartitioncoefficientincomparisonwithsingle-metalsystems

hasalsobeenreported(Jinetal.,1996).Withinthepopulationofsoils,thenaturalvariabilityin

soilcompositionandcompositionofassociatedsoilporewateraresuchastoresultinvariationinKdoverordersofmagnitude,evenforasinglemetal.Forthisreason,anycomprehensivecompilation

ofKdvaluesselectedfromtheliteratureshouldbeexpectedtopresentvaluesthatdefinea

distribution. Infact,foranyparticularmetal,Kddependsontheseandothercharacteristicsofthenaturalmediasystem(soil,sediment,surfacewater),andinanationwideriskassessmentitis

desirabletosamplethenationalpopulationofsuchnaturalmediasystemstoobtainafrequency

distributionofKd.

Unfortunately,thecollectionofnaturalmediasystemschosenforstudybyvariousresea

Partition Coefficients for Metals in Surface Water, Soil and Waste

Documents